Latest AI Research news and updates — updated daily from trusted sources.
Anthropic introduces ‘Claude Corps,’ a fellowship program aimed at training nonprofit organizations to effectively utilize AI technologies. The initiative seeks to democratize access to advanced AI tools and foster social good applications.
Anthropic has reversed a policy that could have limited AI researchers’ ability to develop competing models using Claude. The decision follows backlash from the research community concerned about potential restrictions on AI development.
OpenAI and the Pacific Northwest National Laboratory have introduced DraftNEPABench, a benchmark evaluating how AI coding agents can expedite federal permitting processes. This collaboration aims to enhance efficiency in governmental procedures.
Anthropic launches Claude Fable 5, its most powerful AI model, with restricted access due to safety concerns.
University of Toronto researchers create an AI worm capable of dynamically identifying vulnerabilities and adapting its attack strategies. This development has significant implications for enterprise cybersecurity.
MIT researchers used the game Battleship to train AI agents in strategic thinking and question-asking. This approach enhances AI’s problem-solving capabilities.
An AI agent discovered 21 zero-day vulnerabilities in FFmpeg, highlighting the potential of AI in cybersecurity. This development underscores the need for advanced security measures in software development.
An AI-designed vaccine has successfully passed its first human trial, marking a significant advancement in pandemic preparedness. This development could revolutionize vaccine development processes.
Anthropic has analyzed 832 banned malicious cyber accounts, mapping AI usage in cyber threats to the MITRE ATT&CK framework. This study highlights the evolving role of AI in cyber attacks.
Anthropic analyzes AI-driven cyber threats using MITRE ATT&CK framework.
Anthropic has analyzed 832 banned malicious cyber accounts, revealing that AI use in cyber threats has evolved from basic access work to more sophisticated activities like lateral movement and chained attacks. This mapping to the MITRE ATT&CK framework highlights the growing complexity of AI-driven cyber threats.
China’s DeepSeek AI achieves high performance comparable to leading models using less advanced chips, impressing Silicon Valley technologists. This showcases cost-effective AI development.